RE: The article linked in https://mstdn.fr/@ButterflyOfFire/103171203248328739 ( https://www.devporto.com/fediverse-has-a-discovery-problem-and-thats-not-good/ )
I don't think that having a different kind of discovery from #corpocentric networks means that the #Fediverse's discovery mechanisms are wrong. Having been a #Twitter user for over a decade, they don't have the greatest discovery mechanisms either. They try to fill that in with "who to follow" and trending tags suggestions that are almost always wrong, wrong, wrong.
No universal view is a positive feature, not a negative one. If your group is using the tag #abc123 for something in your local scope (your instance and the instances where your contacts are hosted), it doesn't necessarily collide with another group using #abc123 in their own local scope. This can reduce confusion and conflict. It also means that people can post in whatever languages they desire on instances where said language is the majority and not have posts they can read buried under a multitude of other-language posts.
Many existing Mastodon and Pleroma instances have shared announcement servers (that's not the official name, but I can't recall it right now) that collect public and hashtagged posts and distribute among the other instances using that server. So with judicious choosing of announcement servers, the advantages listed above can be spread over a larger subset of instances and users without as many collision issues as a global view would cause.
No, discovery is not perfect. But let me ask you this: How do you discover e-mail addresses of people you wish to contact? Outside of your organization's address list, you can't just search a directory. You have to ask them or people that already know them. That's probably a better solution than plugging "firstname lastname" or "usual_nick" in a search box anyway.